Diving Beneath the Sea of Stellar Activity: Chromatic Radial Velocities of the Young AU Mic Planetary System

TL;DR

This study refines the masses of the AU Mic planetary system using multi-wavelength radial velocity data and introduces a new Gaussian process toolkit to distinguish stellar activity from planetary signals.

Contribution

It presents a novel wavelength-dependent Gaussian process method for analyzing RV data, improving the ability to detect low-amplitude planetary signals amidst stellar activity.

Findings

Refined mass estimate for AU Mic b: approximately 20.12 Earth masses.

Established an upper limit for AU Mic c's mass at 20.13 Earth masses.

Demonstrated recovery of injected planetary signals with semi-amplitudes as low as 10 m/s.

Abstract

We present updated radial-velocity (RV) analyses of the AU Mic system. AU Mic is a young (22 Myr) early M dwarf known to host two transiting planets - $P_{b}\sim8.46$ days, $R_{b}=4.38_{-0.18}^{+0.18}\ R_{\oplus}$, $P_{c}\sim18.86$ days, $R_{c}=3.51_{-0.16}^{+0.16}\ R_{\oplus}$. With visible RVs from CARMENES-VIS, CHIRON, HARPS, HIRES, {\sc {\textsc{Minerva}}}-Australis, and TRES, as well as near-infrared (NIR) RVs from CARMENES-NIR, CSHELL, IRD, iSHELL, NIRSPEC, and SPIRou, we provide a $5\sigma$ upper limit to the mass of AU Mic c of $M_{c}\leq20.13\ M_{\oplus}$ and present a refined mass of AU Mic b of $M_{b}=20.12_{-1.57}^{+1.72}\ M_{\oplus}$. Used in our analyses is a new RV modeling toolkit to exploit the wavelength dependence of stellar activity present in our RVs via wavelength-dependent Gaussian processes. By obtaining near-simultaneous visible and near-infrared RVs, we also compute the temporal evolution of RV-``color'' and introduce a regressional method to aid in isolating Keplerian from stellar activity signals when modeling RVs in future works. Using a multi-wavelength Gaussian process model, we demonstrate the ability to recover injected planets at $5\sigma$ significance with semi-amplitudes down to $\approx$ 10\,m\,s$^{-1}$ with a known ephemeris, more than an order of magnitude below the stellar activity amplitude. However, we find that the accuracy of the recovered semi-amplitudes is $\sim$50\% for such signals with our model.